Skip to main content
The ESM Atlas is an open-access resource of predicted protein structures and interpretable feature annotations. The ESM Atlas API exposes the same data through REST endpoints at https://biohub.ai/esm/protein/api/v1alpha1. Atlas REST endpoints are public and do not require an API key. The API version is v1alpha1, so interfaces may change without notice. Most endpoints identify a protein by its protein_hash, the MD5 hash that the Atlas uses as the protein’s ID. If you start from a UniProtKB accession, call Resolve UniProt ID first to get the protein_hash.

​
Proteins

​
Batch lookup

Batch lookup retrieves data for many proteins in one request. Small batches return a zip file directly with status 200. Large batches return a job handle with status 202, which you poll until the zip file is ready.

​
Clusters and similarity

​
SAE features

Sparse autoencoder (SAE) features are interpretable signals learned from the ESMC protein language model.

​
Next steps

For step-by-step examples that combine these endpoints, see the Atlas API guides. To query the Atlas from an AI assistant without writing code, use Biohub MCP.